Corrosion parameter adaptive adjustment method and system for wafer wet etching

Abstract

The invention discloses a corrosion parameter adaptive adjustment method and system for wafer wet etching, and relates to the technical field of wafer wet etching, and the method comprises the steps: carrying out the global optimization based on the geometric features of a wafer and wet etching, and obtaining the optimized initial local electrowetting voltage and initial programming type laser scanning path parameters; acquiring a gas-liquid interface curvature radius and a by-product concentration gradient; obtaining an interface transmission impedance value by using a transmission impedance calculation model; inputting the interface transmission impedance value into a corrosion parameter control model, and outputting a local electrowetting voltage adjustment amount and a programmable laser scanning path parameter adjustment amount; and carrying out adaptive adjustment on the initial local electrowetting voltage and the initial programming type laser scanning path parameter. The technical problems that in the prior art, corrosion parameters are difficult to accurately control, real-time performance and adaptability are lacked, corrosion uniformity is poor, the mass transfer effect is not ideal, and the overall process quality is not high are solved.

Description

Technical Field

[0001] This application relates to the field of wafer wet etching technology, specifically to a method and system for adaptive adjustment of etching parameters in wafer wet etching. Background Technology

[0002] With the continuous development of semiconductor manufacturing technology, wafer wet etching processes have been widely used in integrated circuit manufacturing, micro-nano fabrication, and other fields. However, existing technologies make it difficult to precisely control etching parameters during wafer wet etching, resulting in poor etching uniformity, affecting the integrity and performance of the structure, and failing to meet high-precision processing requirements.

[0003] On the other hand, traditional corrosion parameter adjustment methods lack real-time and adaptability. They cannot adjust parameters in a timely manner based on dynamic changes in factors such as the radius of curvature of the gas-liquid interface and the concentration gradient of by-products during wet corrosion. This results in unsatisfactory filling and mass transfer effects of the corrosion solution in high aspect ratio structures, which in turn affects the corrosion rate and overall process quality. Summary of the Invention

[0004] This application provides a method and system for adaptively adjusting corrosion parameters in wafer wet etching, which solves the technical problems of existing technologies, such as difficulty in accurately controlling corrosion parameters, lack of real-time performance and adaptability, resulting in poor corrosion uniformity, unsatisfactory mass transfer effect and low overall process quality.

[0005] The technical solution to the above-mentioned technical problems in this application is as follows: In a first aspect, this application provides a method for adaptively adjusting etching parameters in wafer wet etching, the method comprising: Based on the geometric features of the high aspect ratio structure on the wafer and the process requirements of wet etching, the particle swarm optimization algorithm is used to perform global optimization in the digital twin simulation environment to calculate the optimized initial local electrowetting voltage and initial programmable laser scanning path parameters. During the wet etching process, the interior of the high aspect ratio structure is monitored in situ to obtain the gas-liquid interface curvature radius and by-product concentration gradient in real time. Based on the radius of curvature of the gas-liquid interface and the concentration gradient of the byproducts, the interface transmission impedance value is calculated using a predefined transmission impedance calculation model. The interface transmission impedance value is input into the corrosion parameter control model based on deep reinforcement learning pre-training, and the local electrowetting voltage adjustment amount and the programmable laser scanning path parameter adjustment amount are output. The initial local electrowetting voltage and the initial programmable laser scanning path parameters are adaptively adjusted based on the local electrowetting voltage adjustment amount and the programmable laser scanning path parameter adjustment amount.

[0006] Secondly, this application provides a system for adaptively adjusting etching parameters in wafer wet etching, including: The process pre-optimization module is used to perform global optimization in a digital twin simulation environment based on the geometric features of the high aspect ratio structure on the wafer and the process requirements of wet etching. It calculates the optimized initial local electrowetting voltage and initial programmable laser scanning path parameters by using the particle swarm optimization algorithm in the digital twin simulation environment. The product monitoring module is used to perform in-situ monitoring of the interior of the high aspect ratio structure during the wet corrosion process, and to acquire the gas-liquid interface curvature radius and by-product concentration gradient in real time. The impedance calculation module is used to calculate the interface transmission impedance value based on the radius of curvature of the gas-liquid interface and the concentration gradient of the by-products using a predefined transmission impedance calculation model. The parameter tuning module is used to input the interface transmission impedance value into the corrosion parameter control model based on deep reinforcement learning pre-training, and output the local electrowetting voltage adjustment amount and the programmable laser scanning path parameter adjustment amount; The adjustment execution module adaptively adjusts the initial local electrowetting voltage and the initial programmable laser scanning path parameters based on the adjustment amount of the local electrowetting voltage and the adjustment amount of the programmable laser scanning path parameters.

[0007] This application provides one or more technical solutions, which have at least the following technical effects or advantages: This application provides a method and system for adaptively adjusting corrosion parameters in wafer wet etching. First, a particle swarm optimization algorithm is used in a digital twin simulation environment for global optimization, obtaining optimized initial local electrowetting voltage and initial programmable laser scanning path parameters, laying the foundation for the wet etching process. Second, in-situ monitoring is performed inside the high aspect ratio structure to acquire the gas-liquid interface curvature radius and byproduct concentration gradient in real time, capturing dynamic changes during the process and providing accurate basis for subsequent parameter adjustments. Third, based on the acquired gas-liquid interface curvature radius and byproduct concentration gradient, the interface transmission impedance value is calculated. This value effectively characterizes the comprehensive resistance of the etching solution to filling and mass transfer within the high aspect ratio structure. Then, the interface transmission impedance value is input into a corrosion parameter control model pre-trained based on deep reinforcement learning, outputting adjustments to the local electrowetting voltage and programmable laser scanning path parameters, realizing the function of dynamically adjusting corrosion parameters based on real-time monitoring data. Finally, the initial local electrowetting voltage and initial programmed laser scanning path parameters were adaptively adjusted based on the adjustment amount. In-situ monitoring and parameter adjustments were continuously performed inside the high aspect ratio structure until the wet etching process was completed. This adaptive parameter adjustment mechanism improved the real-time performance and adaptability of corrosion parameter control, effectively improved corrosion uniformity, and enhanced the filling and mass transfer effect of the etching solution within the high aspect ratio structure, thereby improving the structural integrity and performance.

[0008] Through the above technical solutions, this application has formed a complete closed-loop control system from process pre-optimization to real-time monitoring, impedance calculation, parameter tuning and adjustment execution, ensuring the efficient and stable operation of the entire wet corrosion process. Attached Figure Description

[0009] To more clearly illustrate the technical solutions in the embodiments of this application, the accompanying drawings used in the description of the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0010] Figure 1 This is a schematic flowchart of the wafer wet etching corrosion parameter adaptive adjustment method provided in the embodiments of this application; Figure 2 This is a schematic diagram of the structure of the wafer wet etching corrosion parameter adaptive adjustment system provided in the embodiments of this application.

[0011] The components represented by each number in the attached diagram are explained below: Process pre-optimization module 11, product monitoring module 12, impedance calculation module 13, parameter tuning module 14, and adjustment execution module 15. Detailed Implementation

[0012] This application provides a method and system for adaptively adjusting corrosion parameters in wafer wet etching, which addresses the technical problems of existing technologies, such as difficulty in accurately controlling corrosion parameters, lack of real-time performance and adaptability, resulting in poor corrosion uniformity, unsatisfactory mass transfer effect, and low overall process quality.

[0013] The technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this application, and not all embodiments. Based on the embodiments of this application, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this application.

[0014] In the description of this application, the terms "first" and "second" are used for descriptive purposes only and should not be construed as indicating or implying relative importance or implicitly specifying the number of indicated technical features. Thus, a feature defined as "first" or "second" may explicitly or implicitly include one or more of the stated features. In the description of this application, "multiple" means two or more, unless otherwise explicitly specified.

[0015] In the description of this application, the term "for example" is used to mean "used as an example, illustration, or description." Any embodiment described as "for example" in this application is not necessarily to be construed as being more preferred or advantageous than other embodiments. The following description is provided to enable any person skilled in the art to make and use this application. Details are set forth in the following description for purposes of explanation. It should be understood that those skilled in the art will recognize that this application can be made without using these specific details. In other instances, well-known structures and processes will not be described in detail to avoid unnecessarily obscuring the description of this application. Therefore, this application is not intended to be limited to the embodiments shown, but is consistent with the broadest scope of the principles and features disclosed in this application.

[0016] Example 1, as Figure 1 As shown in the embodiments of this application, a method for adaptively adjusting etching parameters in wafer wet etching is provided, including: S10: Based on the geometric features of the high aspect ratio structure on the wafer and the process requirements of wet etching, the particle swarm optimization algorithm is used to perform global optimization in the digital twin simulation environment to calculate the optimized initial local electrowetting voltage and initial programmable laser scanning path parameters. In this embodiment, firstly, the geometric features of the high aspect ratio structure on the wafer include information such as the structure's depth, width, and shape. The initial filling of the etching solution within the high aspect ratio structure is incomplete, resulting in high flow resistance. The process requirements for wet etching involve indicators such as etching rate, etching uniformity, and surface roughness. Particle swarm optimization (PSO) is a swarm intelligence-based optimization algorithm that searches for the optimal solution in the solution space by simulating the collective behavior of flocks of birds or schools of fish. In a digital twin simulation environment, a virtual model highly similar to the actual wafer wet etching process can be created, accurately reflecting various physical phenomena and parameter changes during the etching process.

[0017] Furthermore, a global optimization algorithm is used in the digital twin simulation environment to find an optimal set of initial local electrowetting voltages and initial programmed laser scanning path parameters, taking into account the geometric characteristics of the high aspect ratio structure and the process requirements of wet corrosion. The optimal initial local electrowetting voltages and initial programmed laser scanning path parameters can provide more suitable corrosion conditions for the high aspect ratio structure at the start of the corrosion process, thereby improving the uniformity and quality of corrosion.

[0018] Specifically, each particle in the particle swarm optimization algorithm is considered as a combination of a set of possible initial local electrowetting voltages and initial programmed laser scanning path parameters. Each particle moves in the solution space, adjusting its direction and speed based on its own historical best position and the swarm's historical best position. Through continuous iteration, the particle swarm gradually converges towards the optimal solution, ultimately yielding optimized initial parameters.

[0019] The global optimization method based on digital twin simulation environment and particle swarm optimization algorithm mentioned above avoids the limitations of traditional methods that rely on experience or trial and error to determine initial parameters. It can provide initial conditions for wafer wet etching process in a more scientific and accurate way, and lays a good foundation for subsequent real-time monitoring and parameter adjustment.

[0020] Specifically, step S10 in the method includes: Structures formed on the wafer surface during processing with a ratio of depth to opening width greater than or equal to a preset ratio threshold are identified as high aspect ratio structures on the wafer. Construct a digital twin simulation environment specifically for wet etching of wafers; Extract the geometric feature parameters of the high aspect ratio structure and the process constraint parameters corresponding to the wet corrosion process requirements. The geometric feature parameters include the structure depth and the structure opening width, and the process constraint parameters include the target corrosion rate and the allowable deviation of corrosion uniformity. Within the range of values ​​for local electrowetting voltage and programmable laser scanning path parameters, multiple particles are randomly generated to form an initial particle swarm, wherein the position vector of each particle is encoded as a combination of local electrowetting voltage and programmable laser scanning path parameters. In a digital twin simulation environment, the corrosion process is simulated by combining the local electrowetting voltage corresponding to each particle in the initial particle swarm with the programmable laser scanning path parameters, and simulation results including corrosion depth data, corrosion time data, and structural morphology data are obtained. Calculate the overall process score for each particle based on the simulation results; Based on the comprehensive process score, particle swarm optimization iteration is performed to update the position vector and velocity vector of the particles, and the individual optimal solution of each particle and the global optimal particle of the particle swarm are recorded. When the preset iteration threshold is reached, the particle swarm optimization iteration process is stopped, the position vector of the globally optimal particle is decoded, and the optimized initial local electrowetting voltage and initial programmable laser scanning path parameters are obtained.

[0021] In this embodiment, firstly, high aspect ratio structures on the wafer surface are identified. A preset ratio threshold is set based on process experience and actual needs. When the ratio of the structure depth to the structure opening width is greater than or equal to this threshold, the structure is identified as a high aspect ratio structure, which may cause difficulties in filling the etching solution and uneven mass transfer during wet etching.

[0022] Secondly, a dedicated digital twin simulation environment for wafer wet etching is constructed. This simulation environment mimics the actual wet etching process, taking into account various physical and chemical factors, such as electric field distribution, laser energy transfer, etchant flow, and chemical reactions. By conducting simulations in a virtual environment, extensive testing in actual production can be avoided, saving time and costs.

[0023] Extracting the geometric feature parameters of high aspect ratio structures and the corresponding process constraint parameters for wet corrosion processes is to clarify the optimization objectives and limitations. The structural depth and opening width directly affect the flow and filling of the corrosion solution, while the target corrosion rate and allowable deviation in corrosion uniformity are indicators for evaluating the quality of the corrosion process.

[0024] Furthermore, randomly generating an initial particle swarm is the starting step of the particle swarm optimization algorithm. Multiple particles are randomly generated within the range of local electrowetting voltage and programmable laser scanning path parameters, with each particle representing a set of possible parameter combinations. This random initialization ensures that the algorithm searches within the solution space, avoiding getting trapped in local optima.

[0025] In a digital twin simulation environment, the corrosion process is simulated for each particle with corresponding parameter combinations. The simulation results are obtained, including corrosion depth data, corrosion time data, and structural morphology data, reflecting the corrosion effect under different parameter combinations.

[0026] Then, the overall process score for each particle is calculated based on the simulation results to evaluate the merits of each parameter combination. The overall process score can be obtained by weighting multiple factors such as corrosion rate, corrosion uniformity, and surface roughness. The higher the score, the better the parameter combination meets the process requirements.

[0027] Next, particle swarm optimization iteratively is performed based on the overall process score. In each iteration, particles adjust their movement direction and speed according to their own historical best position and the historical best position of the swarm, continuously moving closer to a better solution. Through continuous iteration, the particle swarm gradually converges to the global optimum.

[0028] Furthermore, the particle swarm optimization iteration process is stopped when a preset iteration threshold is reached. The preset iteration threshold can be adjusted according to the actual situation to balance the computational efficiency and optimization effect of the algorithm. The position vector of the globally optimal particle is decoded to obtain the optimized initial local electrowetting voltage and the initial programmable laser scanning path parameters.

[0029] Furthermore, a high-precision optical profilometer is used to scan and measure the high aspect ratio structure on the wafer to obtain measurement data including the structure opening width, structure depth, sidewall perpendicularity and surface roughness. A three-dimensional geometric model of the high aspect ratio structure was established based on the measurement data. The corrosion rate of wafer materials by etchant under different concentrations and temperatures was tested experimentally. A correlation database of etchant concentration, temperature and corrosion rate was established, and a physicochemical property model of the etchant was constructed based on the correlation database. Design multiple sets of experiments combining different local electrowetting voltages and programmable laser scanning path parameters, record corrosion uniformity data, corrosion rate data and structural integrity detection data corresponding to each set of experiments, and establish a synergistic model of electrowetting effect and laser scanning effect using multiple regression analysis method; The three-dimensional geometric model of the high aspect ratio structure, the physicochemical property model of the corrosion liquid, and the synergistic effect model of electrowetting effect and laser scanning effect are imported into the digital twin platform. Data interaction interfaces between the models are set to complete the construction of a digital twin simulation environment dedicated to wafer wet etching.

[0030] In this embodiment, a high-precision optical profilometer is first used to scan and measure the high aspect ratio structure on the wafer. The high-precision optical profilometer has high resolution and high accuracy, and can accurately acquire key measurement data such as the structure opening width, structure depth, sidewall perpendicularity, and surface roughness.

[0031] Secondly, a three-dimensional geometric model of the high aspect ratio structure is established based on the measurement data. This model can intuitively display the spatial morphology and dimensional characteristics of the high aspect ratio structure, providing a structural basis for subsequent simulation of the corrosion process in a digital twin simulation environment. Through this model, the flow, filling, and mass transfer of the corrosive liquid within the structure can be analyzed more accurately.

[0032] The etching rate of wafer materials under different concentrations and temperatures of etchant was tested experimentally. During the experiments, a correlation database of etchant concentration, temperature, and etching rate was established. Based on this database, a physicochemical property model of the etchant was constructed, accurately simulating the behavior of the etchant under different conditions in a digital twin simulation environment, providing a basis for optimizing the etching process.

[0033] Furthermore, multiple sets of experiments were designed combining different local electrowetting voltages and programmable laser scanning path parameters. During the experiments, corrosion uniformity data, corrosion rate data, and structural integrity detection data were recorded for each set of experiments. A synergistic model of the electrowetting effect and the laser scanning effect was established using multiple regression analysis to quantitatively describe the relationship between the two effects and their impact on the corrosion process.

[0034] Furthermore, after importing the high aspect ratio three-dimensional geometric model, the physicochemical properties model of the etching solution, and the synergistic effect model of electrowetting and laser scanning into the digital twin platform, data interaction interfaces between the models are set up. These interfaces ensure data flow and information sharing between the various models, enabling the digital twin simulation environment to accurately simulate the actual process of wafer wet etching. The construction of this dedicated digital twin simulation environment for wafer wet etching, completed in the above manner, provides a highly realistic and accurate virtual environment for subsequent global optimization using particle swarm optimization algorithms, thus providing more scientific and accurate initial conditions for the wafer wet etching process.

[0035] Specifically, the overall process score for each particle is calculated based on the simulation results, including: Based on the corrosion depth data in the simulation results, the ratio of the difference between the maximum and minimum corrosion depths in different regions of the high aspect ratio structure to the average corrosion depth is calculated to obtain the corrosion uniformity index. Based on the corrosion depth and time data in the simulation results, the ratio of corrosion depth per unit time to the target corrosion rate is calculated to obtain the corrosion rate index. Based on the structural morphology data in the simulation results, the number of depressions, the number of protrusions and the maximum defect size of the structural sidewalls are detected to obtain the structural integrity index. Scoring standards are established for the corrosion uniformity index, the corrosion rate index, and the structural integrity index, respectively, and each index is quantified into dimensionless corrosion uniformity index score value, corrosion rate index score value, and structural integrity index score value. According to preset weights, the corrosion uniformity index score, corrosion rate index score, and structural integrity index score are weighted and summed to calculate the particle's overall process score.

[0036] In this embodiment, firstly, the corrosion uniformity index is calculated based on the corrosion depth data from the simulation results. The corrosion conditions in different regions of a high aspect ratio structure may vary. The uniformity of corrosion is reflected by calculating the ratio of the difference between the maximum and minimum corrosion depths to the average corrosion depth. The smaller this ratio, the more uniform the corrosion. In wet etching processes, corrosion uniformity affects the quality and performance of the structure on the wafer.

[0037] Secondly, the corrosion rate index is calculated based on the corrosion depth and time data from the simulation results. The ratio of corrosion depth to target corrosion rate per unit time can measure the degree of matching between the current corrosion rate and process requirements. If the ratio is close to 1, it indicates that the corrosion rate meets the process requirements; if the ratio deviates significantly from 1, the corrosion parameters need to be adjusted to ensure that the expected corrosion rate is achieved.

[0038] Then, based on the structural morphology data in the simulation results, the number of depressions, protrusions, and the maximum defect size of the structural sidewalls are detected to obtain structural integrity indicators. Defects in the structural sidewalls affect the stability and functionality of the structure; depressions, protrusions, and larger defects may lead to a decline in structural performance.

[0039] Furthermore, scoring criteria were developed for corrosion uniformity, corrosion rate, and structural integrity indices. These indices were quantified into dimensionless scores, allowing for comparison and calculation of indices with different properties on the same scale.

[0040] Finally, the corrosion uniformity index score, corrosion rate index score, and structural integrity index score are weighted and summed according to preset weights to obtain the particle's overall process score. The preset weights are set based on the key requirements of the process and the actual situation; for example, if high corrosion uniformity is required, the weight of the corrosion uniformity index score can be appropriately increased. The calculated overall process score comprehensively and objectively reflects the merits of the combination of local electrowetting voltage and programmable laser scanning path parameters for each particle, providing an accurate evaluation basis for subsequent particle swarm optimization iterations and helping to find optimization parameters that better meet process requirements.

[0041] For example, assume that the preset weights for the corrosion uniformity index, corrosion rate index, and structural integrity index are 0.4, 0.3, and 0.3, respectively. In a certain simulation result, based on the corrosion depth data, the maximum corrosion depth in different regions of the high aspect ratio structure is calculated to be 10 micrometers, the minimum corrosion depth is 6 micrometers, and the average corrosion depth is 8 micrometers. Then, the corrosion uniformity index is (10-6)÷8=0.5.

[0042] When the corrosion uniformity index is 0-0.2, the score is 90-100 points; when it is 0.2-0.4, the score is 70-90 points; when it is 0.4-0.6, the score is 50-70 points; when it is 0.6-0.8, the score is 30-50 points; and when it is 0.8-1, the score is 10-30 points. The corresponding score for this corrosion uniformity index is 60 points.

[0043] Based on the corrosion depth and time data, the corrosion depth per unit time is 2 micrometers / minute, the target corrosion rate is 2.5 micrometers / minute, and the corrosion rate index is 2 ÷ 2.5 = 0.8. If the scoring criteria are: 90-100 points for a corrosion rate index of 0.9-1.1, 70-90 points for 0.7-0.9, 50-70 points for 0.5-0.7, 30-50 points for 0.3-0.5, and 10-30 points for 0.1-0.3, then the corresponding score for this corrosion rate index is 80 points.

[0044] Based on the structural morphology data, three dents and two protrusions were detected on the structural sidewalls, with a maximum defect size of 0.5 micrometers. If the scoring criteria are: 0-2 dents and protrusions with a maximum defect size less than 0.2 micrometers, the score is 90-100 points; 3-5 dents and protrusions with a maximum defect size of 0.2-0.5 micrometers, the score is 70-90 points; 6-8 dents and protrusions with a maximum defect size of 0.5-0.8 micrometers, the score is 50-70 points; and more than 8 dents and protrusions with a maximum defect size greater than 0.8 micrometers, the score for this structural integrity index is 80 points.

[0045] The particle's overall process score is calculated by weighting the results according to the preset weights: 60×0.4+80×0.3+80×0.3=24+24+24=72 points.

[0046] Specifically, step S20 in the method includes: The gas-liquid interface inside the high aspect ratio structure was three-dimensionally scanned using a laser confocal microscopy system to obtain three-dimensional point cloud data of the gas-liquid interface. Extract the three-dimensional point cloud data of the gas-liquid interface in the target area, and fit the local quadratic surface based on the three-dimensional point cloud data of the region using the least squares method. Based on the surface parameters of the local quadratic surface, the average curvature of the target region is calculated, and the reciprocal of the average curvature is used as the radius of curvature of the gas-liquid interface. A specific wavelength of probe laser is emitted into the interior of the high aspect ratio structure using a micro-area laser-induced fluorescence spectroscopy system to excite and collect characteristic fluorescence signals generated by corrosion byproducts. The byproduct concentration gradient is calculated based on the signal intensity distribution of the characteristic fluorescence signal along the depth direction of the high aspect ratio structure.

[0047] In this embodiment, a laser confocal microscopy system is first used to perform a three-dimensional scan of the gas-liquid interface inside the high aspect ratio structure. The laser confocal microscopy system features high resolution and high precision, capturing the characteristics of the gas-liquid interface and obtaining accurate three-dimensional point cloud data, including the spatial position and morphological information of the gas-liquid interface, providing a foundation for subsequent analysis.

[0048] Secondly, regional 3D point cloud data of the target area is extracted from the acquired 3D point cloud data. The selection of the target area is determined based on actual needs and research priorities, and may be a specific region in a high aspect ratio structure prone to uneven corrosion or mass transfer problems. Based on the extracted regional 3D point cloud data, a local quadratic surface is fitted using the least squares method. The least squares method is a commonly used mathematical method that can find a quadratic surface that best fits the point cloud data, minimizing the sum of squared distances from the point cloud data to this surface. The local quadratic surface can well approximate the shape of the gas-liquid interface in the target area.

[0049] Then, the average curvature of the target region is calculated based on the surface parameters of the local quadratic surface. The average curvature reflects the degree of curvature of the surface. The reciprocal of the average curvature is used as the radius of curvature of the gas-liquid interface, which is used to analyze the influence of the geometric properties of the gas-liquid interface on the wet corrosion process.

[0050] For example, after performing a three-dimensional scan of the gas-liquid interface inside a high aspect ratio structure using a laser confocal microscopy system, the parameters of the local quadratic surface are obtained by least squares fitting. Assuming the calculated average curvature of this local quadratic surface is 0.05, then the radius of curvature of the gas-liquid interface is 1 ÷ 0.05 = 20 micrometers.

[0051] Simultaneously, a micro-area laser-induced fluorescence spectroscopy system is used to emit a probe laser of a specific wavelength into the interior of the high aspect ratio structure. The selection of the specific wavelength is determined based on the fluorescence characteristics of the corrosion byproducts, enabling them to generate characteristic fluorescence signals. By acquiring these characteristic fluorescence signals and analyzing their signal intensity distribution along the depth direction of the high aspect ratio structure, the byproduct concentration gradient is obtained. This byproduct concentration gradient reflects the distribution of corrosion byproducts within the structure. In wet corrosion processes, an excessively large byproduct concentration gradient may lead to problems such as uneven corrosion or a decrease in the corrosion rate.

[0052] S30: Based on the radius of curvature of the gas-liquid interface and the concentration gradient of the byproduct, the interface transmission impedance value is calculated using a predefined transmission impedance calculation model. In this embodiment, the predefined transmission impedance calculation model is established based on the physical properties of the gas-liquid interface and the transport laws of corrosion byproducts. The radius of curvature of the gas-liquid interface reflects the geometry of the interface, and different radii of curvature affect the flow and mass transfer of the liquid at the interface. The byproduct concentration gradient reflects the distribution differences of corrosion byproducts within the high aspect ratio structure, and these distribution differences lead to changes in the driving force for mass transport.

[0053] In the model, the radius of curvature of the gas-liquid interface is related to factors such as the surface tension of the gas-liquid interface and the viscosity of the liquid. A smaller radius of curvature indicates a greater degree of interface tortuosity, which may increase the flow resistance of the liquid at that interface, thus affecting material transport. The byproduct concentration gradient is related to the diffusion coefficient and diffusion path of corrosion byproducts. A larger concentration gradient provides a greater driving force for the diffusion of corrosion byproducts, but it may also be limited by the internal space of the structure.

[0054] By using the radius of curvature of the gas-liquid interface and the concentration gradient of byproducts as input parameters and substituting them into a predefined transmission impedance calculation model, the interface transmission impedance value is obtained. This value reflects the ease or difficulty of material transport at the gas-liquid interface. In the wet etching process of wafers, the interface transmission impedance value directly affects the material exchange rate between the etching solution and the wafer material, thereby affecting the uniformity and rate of etching.

[0055] For example, when the radius of curvature of the gas-liquid interface is small and the concentration gradient of byproducts is large, the interfacial transport impedance may be large, indicating that the material transport is significantly hindered, which may lead to a decrease in the local corrosion rate and uneven corrosion. Conversely, when the radius of curvature of the gas-liquid interface is large and the concentration gradient of byproducts is small, the interfacial transport impedance may be small, the material transport is relatively smooth, and the corrosion process may be more uniform and efficient.

[0056] Specifically, step S30 in the method includes: The real-time acquired gas-liquid interface curvature radius and byproduct concentration gradient are input into a predefined transmission impedance calculation model to calculate the interface transmission impedance value. The interface transmission impedance value is used to characterize the comprehensive resistance of the corrosive liquid to filling and transferring mass within the high aspect ratio structure.

[0057] In this embodiment, the radius of curvature of the gas-liquid interface and the concentration gradient of by-products are first acquired in real time. As the wet etching process of the wafer progresses, the geometry of the gas-liquid interface and the distribution of corrosion by-products will continuously change.

[0058] The real-time acquired radius of curvature of the gas-liquid interface and the byproduct concentration gradient are input into a predefined transmission impedance calculation model. This model is based on extensive experimental data and theoretical analysis, comprehensively considering the physical properties of the gas-liquid interface and the transport characteristics of corrosion byproducts. In the model, the radius of curvature of the gas-liquid interface is related to factors such as the surface tension and viscosity of the liquid, collectively influencing the flow and mass transfer of the liquid at the interface. The byproduct concentration gradient is related to the diffusion coefficient and diffusion path of the corrosion byproducts, reflecting the distribution differences of corrosion byproducts within the high aspect ratio structure.

[0059] The predefined transmission impedance calculation model is: Interface transmission impedance Z = α / R + β × |▽C|. R represents the radius of curvature of the gas-liquid interface, ▽C represents the byproduct concentration gradient, and α and β are weighting coefficients determined based on the geometric characteristics of the high aspect ratio structure and the requirements of the wet corrosion process.

[0060] Furthermore, through model calculations, the interfacial transport impedance value was obtained, which characterizes the overall resistance of the etchant to filling and transferring mass within the high aspect ratio structure. During wafer wet etching, the magnitude of the interfacial transport impedance value affects the mass exchange rate between the etchant and the wafer material. A large interfacial transport impedance value indicates significant obstruction to mass transport, making it difficult for the etchant to smoothly fill and transfer mass within the high aspect ratio structure, leading to a reduced local etching rate and uneven etching. Conversely, a small interfacial transport impedance value allows for relatively smooth mass transport, potentially resulting in a more uniform and efficient etching process.

[0061] By calculating the interfacial transmission impedance in real time, potential problems during the corrosion process can be identified. For example, a sudden increase in the interfacial transmission impedance indicates an unfavorable change in the morphology of the gas-liquid interface or an abnormal distribution of corrosion byproducts. In such cases, corrosion process parameters, such as the concentration and temperature of the corrosive solution, can be adjusted promptly to reduce the interfacial transmission impedance and ensure the smooth progress of the corrosion process.

[0062] S40: Input the interface transmission impedance value into the corrosion parameter control model based on deep reinforcement learning pre-training, and output the local electrowetting voltage adjustment amount and the programmable laser scanning path parameter adjustment amount; In this embodiment, a corrosion parameter control model pre-trained based on deep reinforcement learning learns the complex relationship between the interface transmission impedance value, the local electrowetting voltage, and the programmable laser scanning path parameters to achieve intelligent adjustment of corrosion parameters. When the interface transmission impedance value is input into the pre-trained model, the model outputs the adjustment amount of the local electrowetting voltage and the adjustment amount of the programmable laser scanning path parameters according to the mapping relationship it has learned.

[0063] Adjusting the local electrowetting voltage can alter the wettability and flowability of the liquid within a high aspect ratio structure, thereby affecting the contact between the etchant and the wafer material and the mass transfer rate. For example, appropriately increasing the local electrowetting voltage may allow the etchant to better fill the high aspect ratio structure, improving mass transfer efficiency and thus enhancing the uniformity and rate of etching. Meanwhile, adjusting the programmable laser scanning path parameters allows for precise control of the etching process. Different laser scanning paths affect factors such as temperature distribution and chemical reaction rates within the etched region; by adjusting these parameters, the etching effect can be optimized.

[0064] Specifically, in practical applications, the output adjustments for local electrowetting voltage and programmable laser scanning path parameters are based on the model's analysis and judgment of the current interface transmission impedance. If the interface transmission impedance is high, the model may output an increase in the local electrowetting voltage adjustment while simultaneously adjusting the programmable laser scanning path parameters to improve material transport and reduce interface transmission impedance. Conversely, if the interface transmission impedance is low, the model may appropriately reduce the local electrowetting voltage or fine-tune the programmable laser scanning path parameters to maintain the stability and efficiency of the corrosion process.

[0065] In this way, the corrosion parameter control model based on deep reinforcement learning pre-training can dynamically adjust the local electrowetting voltage and programmable laser scanning path parameters according to the real-time interface transmission impedance value, thereby achieving adaptive control of the wafer wet etching process, improving the quality and stability of the etching process, and meeting the wafer manufacturing requirements under different process requirements.

[0066] Specifically, step S40 in the method includes: Retrieve the pre-trained corrosion parameter control model; The real-time acquired interface transmission impedance value and the historical sequence data of the interface transmission impedance value are combined to form a characterization of the current process status. The current process state is input into the corrosion parameter control model, and the local electrowetting voltage adjustment and the programmable laser scanning path parameter adjustment are output.

[0067] In this embodiment, firstly, a corrosion parameter control model pre-trained based on deep reinforcement learning is retrieved. This model learns the complex mapping relationship between the interface transmission impedance value, the local electrowetting voltage, and the programmable laser scanning path parameters.

[0068] Secondly, the interface transmission impedance value and its historical sequence data are acquired in real time. During the wafer wet etching process, the interface transmission impedance value is constantly changing. The historical sequence data contains information on the interface transmission impedance value at various previous moments, reflecting the dynamic change trend of the etching process. Combining the real-time acquired interface transmission impedance value with the historical sequence data constitutes a characterization of the current process status.

[0069] Then, the current process state characterization is input into the retrieved corrosion parameter control model. The model performs in-depth analysis and processing of the input information, and calculates and outputs the local electrowetting voltage adjustment and the programmable laser scanning path parameter adjustment based on its learned knowledge and rules. By adjusting the local electrowetting voltage and the programmable laser scanning path parameters, the material transport conditions during the corrosion process are improved, thereby increasing the uniformity and rate of corrosion and ensuring the quality and stability of the wafer wet etching process.

[0070] The construction process of the corrosion parameter control model includes: Based on the aforementioned digital twin simulation environment, a simulation training environment is established with the interface transmission impedance value as the state input and the local electrowetting voltage adjustment amount and the programmable laser scanning path parameter adjustment amount as the control action output. Based on the deep reinforcement learning framework, a corrosion parameter control model including a policy network and a value network is constructed. A reward function is designed with the optimization objective of minimizing the long-term accumulation and instantaneous fluctuation of the interface transmission impedance value; In the simulation training environment, the corrosion parameter control model is iteratively trained using a deep reinforcement learning algorithm until the control strategy converges, thus obtaining a pre-trained corrosion parameter control model.

[0071] In this embodiment, firstly, a simulation training environment is established based on a digital twin simulation environment. The digital twin simulation environment can highly reproduce the real process of wet etching on wafers, using the interface transmission impedance value as the state input. This is because the interface transmission impedance value reflects the comprehensive resistance of the etchant to filling and transferring mass within a high aspect ratio structure. The local electrowetting voltage adjustment and the programmable laser scanning path parameter adjustment are used as control action outputs; adjusting these two parameters directly affects the etching effect and uniformity. Based on the above settings, a simulation training environment is constructed that can simulate the changes in the etching process under different etching parameter adjustments.

[0072] Secondly, a corrosion parameter control model is constructed based on a deep reinforcement learning framework. Deep reinforcement learning frameworks possess powerful learning and decision-making capabilities, and a corrosion parameter control model comprising a policy network and a value network is built within this framework. The policy network generates control actions based on the current state input, i.e., the interface transmission impedance value, which includes adjustments to the local electrowetting voltage and programmable laser scanning path parameters.

[0073] Then, a reward function is designed with the optimization objective of minimizing the long-term accumulation and instantaneous fluctuation of the interface transmission impedance. The long-term accumulated interface transmission impedance reflects the overall resistance to material transport during the entire corrosion process, while the instantaneous fluctuation reflects the stability of the corrosion process. By designing the reward function, the model can obtain a higher reward when the control action minimizes both the long-term accumulation and instantaneous fluctuation of the interface transmission impedance; conversely, it obtains a lower reward.

[0074] For example, the steps for constructing and training a corrosion parameter control model based on a deep reinforcement learning framework are as follows: First, data preparation: Based on the digital twin simulation environment, the interface transmission impedance value and the historical sequence data of the interface transmission impedance value are collected and combined to form a characterization of the current process status.

[0075] Secondly, the model is constructed, taking the current process state as input and the local electrowetting voltage adjustment and the programmable laser scanning path parameter adjustment as output. The number of nodes in the input layer equals the dimension of the input features. For example, if the current process state has 3 features, the input layer contains 3 nodes. 1-3 hidden layers are set, with the number of nodes in each layer adjusted experimentally, such as 64 or 32. The activation function is ReLU. The number of nodes in the output layer equals the number of local electrowetting voltage adjustments and programmable laser scanning path parameter adjustments. For example, if the output time requires 1 node, the output layer generally does not use an activation function and directly outputs continuous values.

[0076] Then, the model is trained with the optimization objective of minimizing the long-term accumulation and instantaneous fluctuation of the interface transmission impedance value. A reward function is designed: R = -α × (ΣZ + β × σ(Z)), where R is the reward value, ΣZ represents the long-term accumulation of the interface transmission impedance value, σ(Z) represents the instantaneous fluctuation of the interface transmission impedance value, and α and β are weight coefficients adjusted according to the actual situation. In each training iteration, the Adam optimizer and the mean squared error (MSE) loss function are used to construct the training framework. The batch size is set to 32, the total number of training rounds is 500, and an early stopping mechanism (patience = 5) is introduced. When the validation set loss does not decrease for 5 consecutive rounds, the training process is automatically terminated, resulting in the trained corrosion parameter control model.

[0077] Finally, the erosion parameter control model was iteratively trained using a deep reinforcement learning algorithm in a simulation training environment. During training, the model continuously generates control actions based on the current state input, then obtains corresponding rewards according to the reward function, and adjusts the parameters of the policy network and value network based on the reward feedback. After multiple iterations of training, the control policy converges, meaning the model can stably output the optimal control action. At this point, the pre-trained erosion parameter control model is obtained.

[0078] The pre-trained model can quickly and accurately adjust the local electrowetting voltage and programmable laser scanning path parameters based on the real-time interface transmission impedance value during actual wafer wet etching process, so as to ensure the uniformity and efficiency of the etching process.

[0079] S50: Adaptively adjust the initial local electrowetting voltage and the initial programmable laser scanning path parameters according to the local electrowetting voltage adjustment amount and the programmable laser scanning path parameter adjustment amount.

[0080] In this embodiment, after obtaining the local electrowetting voltage adjustment amount and the programmable laser scanning path parameter adjustment amount, the initial local electrowetting voltage and the initial programmable laser scanning path parameters are adaptively adjusted according to the adjustment amount. For the adjustment of the initial local electrowetting voltage, the calculated local electrowetting voltage adjustment amount is used. If the adjustment amount is positive, it indicates that the local electrowetting voltage needs to be increased to enhance the wettability and flowability of the etchant in the high aspect ratio structure, allowing the etchant to better contact the wafer material, increasing the mass exchange rate, and thus improving the etching effect. For example, if the original local electrowetting voltage is 5V and the adjustment amount is 1V, then the adjusted local electrowetting voltage becomes 6V. Conversely, if the adjustment amount is negative, the local electrowetting voltage needs to be decreased.

[0081] Adjusting the initial programmed laser scanning path parameters requires modifying the laser scanning path, speed, and frequency based on the adjustment amount. Different laser scanning path parameters will affect the temperature distribution and chemical reaction rate in the etched area. For example, if the adjustment indicates a need to change the laser scanning path, the laser scanning trajectory needs to be replanned to ensure more uniform laser illumination of the wafer surface, thereby optimizing the etching process. If the original laser scanning speed is 10 mm / s, and the adjustment requires an increase of 2 mm / s, then the adjusted laser scanning speed will become 12 mm / s.

[0082] By adaptively adjusting the initial local electrowetting voltage and the initial programmed laser scanning path parameters, the wafer wet etching process can be made to better meet the expected requirements.

[0083] Specifically, step S50 in the method includes: The local electrowetting voltage adjustment is added to the initial local electrowetting voltage to obtain the updated local electrowetting voltage; The adjusted amount of the programmable laser scanning path parameters is added to the initial programmable laser scanning path parameters to obtain the updated programmable laser scanning path parameters; Based on the updated local electrowetting voltage, the electrowetting voltage applied to the bottom of the high aspect ratio structure is adjusted; The laser scanning path acting on the sidewall of the high aspect ratio structure is adjusted according to the updated programmable laser scanning path parameters. The interior of the high aspect ratio structure is continuously monitored in situ, and the interface transmission impedance value is calculated, new local electrowetting voltage adjustment amount and new programmable laser scanning path parameter adjustment amount are obtained until the wet etching process is completed.

[0084] In this embodiment, the local electrowetting voltage adjustment is first added to the initial local electrowetting voltage to obtain an updated local electrowetting voltage value. The updated local electrowetting voltage affects the state of the liquid within the high aspect ratio structure. When the updated local electrowetting voltage increases, the wettability and fluidity of the liquid are enhanced, allowing the etchant to more fully fill the structure's interior, increasing the contact area with the wafer material, and improving the mass exchange rate, thereby promoting the corrosion reaction. Conversely, if the updated local electrowetting voltage decreases, the relevant properties of the liquid weaken, and the rate and effectiveness of the corrosion reaction change accordingly.

[0085] Secondly, the adjusted programmable laser scanning path parameters are added to the initial programmable laser scanning path parameters to obtain the updated programmable laser scanning path parameters. The updated parameters determine the specific characteristics of the laser scan. Changes in the laser scanning path directly affect the temperature distribution within the corrosion zone. For example, a denser scanning path may increase local temperatures and accelerate the chemical reaction rate; while a sparser scanning path may result in a more uniform temperature distribution, but the reaction rate may be relatively slower. Simultaneously, changes in parameters such as laser scanning speed and frequency also have different effects on the corrosion process.

[0086] For example, in a real-world wafer wet etching case, the initial local electrowetting voltage is set to 4V, and the initial programmable laser scanning path parameters include a scanning speed of 8mm / s and a scanning frequency of 10Hz. After model calculation in step S40, the local electrowetting voltage adjustment is 0.5V, and the programmable laser scanning path parameter adjustments include a scanning speed adjustment of 1mm / s and a scanning frequency adjustment of -2Hz.

[0087] Following step S50, the updated local electrowetting voltage is first calculated. The local electrowetting voltage adjustment of 0.5V is added to the initial local electrowetting voltage of 4V to obtain an updated local electrowetting voltage of 4.5V. Then, the electrowetting voltage applied to the bottom of the high aspect ratio structure is adjusted to 4.5V. At this point, the wettability and fluidity of the etchant within the high aspect ratio structure are enhanced, allowing for better contact with the wafer material and increasing the mass exchange rate.

[0088] Then, the updated programmable laser scanning path parameters are calculated. The scanning speed adjustment amount of 1 mm / s is added to the initial scanning speed of 8 mm / s to obtain the updated scanning speed of 9 mm / s. The scanning frequency adjustment amount of -2 Hz is added to the initial scanning frequency of 10 Hz to obtain the updated scanning frequency of 8 Hz.

[0089] Subsequently, based on the updated programmable laser scanning path parameters, the laser scanning path acting on the sidewall of the high aspect ratio structure was adjusted, resulting in a scanning speed of 9 mm / s and a scanning frequency of 8 Hz. The increased scanning speed may allow the laser to cover a larger area per unit time, accelerating heat transfer and distribution, thus affecting the temperature change in the corroded area. Conversely, the decreased scanning frequency may increase the interval between laser pulses, reducing the stimulation frequency for the chemical reaction, but it may also make the reaction more stable.

[0090] Then, based on the updated local electrowetting voltage, the electrowetting voltage applied to the bottom of the high aspect ratio structure is adjusted.

[0091] Next, based on the updated programmed laser scanning path parameters, the laser scanning path acting on the sidewalls of the high aspect ratio structure is adjusted. This requires resetting and calibrating the laser scanning equipment. For example, parameters such as the laser emission angle, frequency, and speed are changed to achieve the new scanning path. During the adjustment process, it is crucial to ensure that the laser accurately acts on the structural sidewalls to avoid scanning deviations.

[0092] After adjusting the electrowetting voltage and laser scanning path, continuous in-situ monitoring of the interior of the high aspect ratio structure is required. For example, monitoring techniques such as optical microscopy and electron microscopy can be used to observe the corrosion process inside the structure in real time. Simultaneously, the interface transmission impedance value is calculated, and based on the new interface transmission impedance value, new local electrowetting voltage adjustments and new programmable laser scanning path parameter adjustments are obtained. This cyclical adjustment process continues until the wet etching process is complete, ensuring the uniformity and efficiency of the entire etching process and meeting the high-quality requirements of wafer manufacturing.

[0093] In summary, compared to existing technologies, this application achieves adaptive control of the wafer wet etching process by constructing a corrosion parameter control model based on deep reinforcement learning and dynamically adjusting the local electrowetting voltage and programmable laser scanning path parameters according to the real-time interface transmission impedance value. This avoids the problems of accurately grasping corrosion parameter changes and the inability to adjust them in real time in traditional methods.

[0094] In summary, the embodiments of this application have at least the following technical effects: This application provides a method for adaptively adjusting corrosion parameters in wafer wet etching. First, a particle swarm optimization algorithm is used in a digital twin simulation environment for global optimization, obtaining optimized initial local electrowetting voltage and initial programmable laser scanning path parameters, laying the foundation for the wet etching process. Second, in-situ monitoring is performed inside the high aspect ratio structure to acquire the gas-liquid interface curvature radius and byproduct concentration gradient in real time, capturing dynamic changes during the process and providing accurate basis for subsequent parameter adjustments. Third, based on the acquired gas-liquid interface curvature radius and byproduct concentration gradient, the interface transmission impedance value is calculated. This value effectively characterizes the comprehensive resistance of the etchant to filling and mass transfer within the high aspect ratio structure. Then, the interface transmission impedance value is input into a corrosion parameter control model pre-trained based on deep reinforcement learning, outputting adjustments to the local electrowetting voltage and programmable laser scanning path parameters, realizing the function of dynamically adjusting corrosion parameters based on real-time monitoring data. Finally, the initial local electrowetting voltage and initial programmed laser scanning path parameters are adaptively adjusted according to the adjustment amount. In-situ monitoring and parameter adjustments are continuously performed inside the high aspect ratio structure until the wet etching process is completed. This adaptive parameter adjustment mechanism improves the real-time performance and adaptability of corrosion parameter control, effectively improves corrosion uniformity, and enhances the filling and mass transfer effect of the etching solution within the high aspect ratio structure, thereby improving the structural integrity and performance. Through the above technical solution, this application forms a complete closed-loop control system from process pre-optimization to real-time monitoring, impedance calculation, parameter tuning, and adjustment execution, ensuring the efficient and stable operation of the entire wet etching process.

[0095] Example 2, as Figure 2 As shown, based on the same inventive concept as the wafer wet etching parameter adaptive adjustment method provided in Embodiment 1, this application also provides a wafer wet etching parameter adaptive adjustment system, including: The process pre-optimization module 11 is used to perform global optimization in a digital twin simulation environment based on the geometric features of the high aspect ratio structure on the wafer and the process requirements of wet etching, and to calculate the optimized initial local electrowetting voltage and initial programmable laser scanning path parameters. The product monitoring module 12 is used to perform in-situ monitoring of the interior of the high aspect ratio structure during the wet corrosion process, and to obtain the gas-liquid interface curvature radius and by-product concentration gradient in real time. Impedance calculation module 13 is used to calculate the interface transmission impedance value based on the radius of curvature of the gas-liquid interface and the concentration gradient of the by-products using a predefined transmission impedance calculation model. The parameter tuning module 14 is used to input the interface transmission impedance value into the corrosion parameter control model based on deep reinforcement learning pre-training, and output the local electrowetting voltage adjustment amount and the programmable laser scanning path parameter adjustment amount. The adjustment execution module 15 adaptively adjusts the initial local electrowetting voltage and the initial programmable laser scanning path parameters based on the local electrowetting voltage adjustment amount and the programmable laser scanning path parameter adjustment amount.

[0096] In one embodiment, the process pre-optimization module 11 is specifically used for: Structures formed on the wafer surface during processing with a ratio of depth to opening width greater than or equal to a preset ratio threshold are identified as high aspect ratio structures on the wafer. Construct a digital twin simulation environment specifically for wet etching of wafers; Extract the geometric feature parameters of the high aspect ratio structure and the process constraint parameters corresponding to the wet corrosion process requirements. The geometric feature parameters include the structure depth and the structure opening width, and the process constraint parameters include the target corrosion rate and the allowable deviation of corrosion uniformity. Within the range of values ​​for local electrowetting voltage and programmable laser scanning path parameters, multiple particles are randomly generated to form an initial particle swarm, wherein the position vector of each particle is encoded as a combination of local electrowetting voltage and programmable laser scanning path parameters. In a digital twin simulation environment, the corrosion process is simulated by combining the local electrowetting voltage corresponding to each particle in the initial particle swarm with the programmable laser scanning path parameters, and simulation results including corrosion depth data, corrosion time data, and structural morphology data are obtained. Calculate the overall process score for each particle based on the simulation results; Based on the comprehensive process score, particle swarm optimization iteration is performed to update the position vector and velocity vector of the particles, and the individual optimal solution of each particle and the global optimal particle of the particle swarm are recorded. When the preset iteration threshold is reached, the particle swarm optimization iteration process is stopped, the position vector of the globally optimal particle is decoded, and the optimized initial local electrowetting voltage and initial programmable laser scanning path parameters are obtained.

[0097] Furthermore, in one embodiment of the application, a high-precision optical profilometer is used to scan and measure the high aspect ratio structure on the wafer to obtain measurement data including the structure opening width, structure depth, sidewall perpendicularity and surface roughness. A three-dimensional geometric model of the high aspect ratio structure was established based on the measurement data. The corrosion rate of wafer materials by etchant under different concentrations and temperatures was tested experimentally. A correlation database of etchant concentration, temperature and corrosion rate was established, and a physicochemical property model of the etchant was constructed based on the correlation database. Design multiple sets of experiments combining different local electrowetting voltages and programmable laser scanning path parameters, record corrosion uniformity data, corrosion rate data and structural integrity detection data corresponding to each set of experiments, and establish a synergistic model of electrowetting effect and laser scanning effect using multiple regression analysis method; The three-dimensional geometric model of the high aspect ratio structure, the physicochemical property model of the corrosion liquid, and the synergistic effect model of electrowetting effect and laser scanning effect are imported into the digital twin platform. Data interaction interfaces between the models are set to complete the construction of a digital twin simulation environment dedicated to wafer wet etching.

[0098] Furthermore, in one embodiment of the application, the overall process score for each particle is calculated based on the simulation results, including: Based on the corrosion depth data in the simulation results, the ratio of the difference between the maximum and minimum corrosion depths in different regions of the high aspect ratio structure to the average corrosion depth is calculated to obtain the corrosion uniformity index. Based on the corrosion depth and time data in the simulation results, the ratio of corrosion depth per unit time to the target corrosion rate is calculated to obtain the corrosion rate index. Based on the structural morphology data in the simulation results, the number of depressions, the number of protrusions and the maximum defect size of the structural sidewalls are detected to obtain the structural integrity index. Scoring standards are established for the corrosion uniformity index, the corrosion rate index, and the structural integrity index, respectively, and each index is quantified into dimensionless corrosion uniformity index score value, corrosion rate index score value, and structural integrity index score value. According to preset weights, the corrosion uniformity index score, corrosion rate index score, and structural integrity index score are weighted and summed to calculate the particle's overall process score.

[0099] In one embodiment, the product monitoring module 12 is specifically used for: The gas-liquid interface inside the high aspect ratio structure was three-dimensionally scanned using a laser confocal microscopy system to obtain three-dimensional point cloud data of the gas-liquid interface. Extract the three-dimensional point cloud data of the gas-liquid interface in the target area, and fit the local quadratic surface based on the three-dimensional point cloud data of the region using the least squares method. Based on the surface parameters of the local quadratic surface, the average curvature of the target region is calculated, and the reciprocal of the average curvature is used as the radius of curvature of the gas-liquid interface. A specific wavelength of probe laser is emitted into the interior of the high aspect ratio structure using a micro-area laser-induced fluorescence spectroscopy system to excite and collect characteristic fluorescence signals generated by corrosion byproducts. The byproduct concentration gradient is calculated based on the signal intensity distribution of the characteristic fluorescence signal along the depth direction of the high aspect ratio structure.

[0100] In one embodiment, the impedance calculation module 13 is specifically used for: The real-time acquired gas-liquid interface curvature radius and byproduct concentration gradient are input into a predefined transmission impedance calculation model to calculate the interface transmission impedance value. The interface transmission impedance value is used to characterize the comprehensive resistance of the corrosive liquid to filling and transferring mass within the high aspect ratio structure.

[0101] In one embodiment, the parameter tuning module 14 is specifically used for: Retrieve the pre-trained corrosion parameter control model; The real-time acquired interface transmission impedance value and the historical sequence data of the interface transmission impedance value are combined to form a characterization of the current process status. The current process state is input into the corrosion parameter control model, and the local electrowetting voltage adjustment and the programmable laser scanning path parameter adjustment are output.

[0102] Furthermore, in one embodiment, the process of constructing the corrosion parameter control model includes: Based on the aforementioned digital twin simulation environment, a simulation training environment is established with the interface transmission impedance value as the state input and the local electrowetting voltage adjustment amount and the programmable laser scanning path parameter adjustment amount as the control action output. Based on the deep reinforcement learning framework, a corrosion parameter control model including a policy network and a value network is constructed. A reward function is designed with the optimization objective of minimizing the long-term accumulation and instantaneous fluctuation of the interface transmission impedance value; In the simulation training environment, the corrosion parameter control model is iteratively trained using a deep reinforcement learning algorithm until the control strategy converges, thus obtaining a pre-trained corrosion parameter control model.

[0103] In one embodiment, the adjustment execution module 15 is specifically used for: The local electrowetting voltage adjustment is added to the initial local electrowetting voltage to obtain the updated local electrowetting voltage; The adjusted amount of the programmable laser scanning path parameters is added to the initial programmable laser scanning path parameters to obtain the updated programmable laser scanning path parameters; Based on the updated local electrowetting voltage, the electrowetting voltage applied to the bottom of the high aspect ratio structure is adjusted; The laser scanning path acting on the sidewall of the high aspect ratio structure is adjusted according to the updated programmable laser scanning path parameters. The interior of the high aspect ratio structure is continuously monitored in situ, and the interface transmission impedance value is calculated, new local electrowetting voltage adjustment amount and new programmable laser scanning path parameter adjustment amount are obtained until the wet etching process is completed.

[0104] It should be noted that the order of the embodiments described above is merely for descriptive purposes and does not represent the superiority or inferiority of the embodiments. Furthermore, the above description focuses on specific embodiments of this specification. Additionally, the processes depicted in the accompanying drawings do not necessarily require a specific or sequential order to achieve the desired results. In some implementations, multitasking and parallel processing are possible or may be advantageous.

[0105] The above description is only a preferred embodiment of this application and is not intended to limit this application. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of this application should be included within the protection scope of this application.

[0106] This specification and accompanying drawings are merely illustrative examples of this application and are intended to cover any and all modifications, variations, combinations, or equivalents within the scope of this application. Clearly, those skilled in the art can make various alterations and modifications to this application without departing from its scope. Therefore, if such modifications and modifications fall within the scope of this application and its equivalents, this application intends to include such modifications and modifications.

Claims

1. A method for adaptively adjusting corrosion parameters in wet etching of wafers, characterized in that, The method includes: Based on the geometric features of the high aspect ratio structure on the wafer and the process requirements of wet etching, the particle swarm optimization algorithm is used to perform global optimization in the digital twin simulation environment to calculate the optimized initial local electrowetting voltage and initial programmable laser scanning path parameters. During the wet etching process, the interior of the high aspect ratio structure is monitored in situ to obtain the gas-liquid interface curvature radius and by-product concentration gradient in real time. Based on the radius of curvature of the gas-liquid interface and the concentration gradient of the byproducts, the interface transmission impedance value is calculated using a predefined transmission impedance calculation model. The interface transmission impedance value is input into the corrosion parameter control model based on deep reinforcement learning pre-training, and the local electrowetting voltage adjustment amount and the programmable laser scanning path parameter adjustment amount are output. The initial local electrowetting voltage and the initial programmable laser scanning path parameters are adaptively adjusted based on the local electrowetting voltage adjustment amount and the programmable laser scanning path parameter adjustment amount.

2. The method for adaptive adjustment of etching parameters in wafer wet etching according to claim 1, characterized in that, Based on the geometric features of the high aspect ratio structure on the wafer and the process requirements of wet etching, a particle swarm optimization algorithm is used in a digital twin simulation environment for global optimization. The optimized initial local electrowetting voltage and initial programmable laser scanning path parameters are calculated, including: Structures formed on the wafer surface during processing with a ratio of depth to opening width greater than or equal to a preset ratio threshold are identified as high aspect ratio structures on the wafer. Construct a digital twin simulation environment specifically for wet etching of wafers; Extract the geometric feature parameters of the high aspect ratio structure and the process constraint parameters corresponding to the wet corrosion process requirements. The geometric feature parameters include the structure depth and the structure opening width, and the process constraint parameters include the target corrosion rate and the allowable deviation of corrosion uniformity. Within the range of values ​​for local electrowetting voltage and programmable laser scanning path parameters, multiple particles are randomly generated to form an initial particle swarm, wherein the position vector of each particle is encoded as a combination of local electrowetting voltage and programmable laser scanning path parameters. In a digital twin simulation environment, the corrosion process is simulated by combining the local electrowetting voltage corresponding to each particle in the initial particle swarm with the programmable laser scanning path parameters, and simulation results including corrosion depth data, corrosion time data, and structural morphology data are obtained. Calculate the overall process score for each particle based on the simulation results; Based on the comprehensive process score, particle swarm optimization iteration is performed to update the position vector and velocity vector of the particles, and the individual optimal solution of each particle and the global optimal particle of the particle swarm are recorded. When the preset iteration threshold is reached, the particle swarm optimization iteration process is stopped, the position vector of the globally optimal particle is decoded, and the optimized initial local electrowetting voltage and initial programmable laser scanning path parameters are obtained.

3. The method for adaptive adjustment of etching parameters in wafer wet etching according to claim 2, characterized in that, Constructing a dedicated digital twin simulation environment for wafer wet etching, including: A high-precision optical profilometer is used to scan and measure the high aspect ratio structure on the wafer to obtain measurement data including the structure opening width, structure depth, sidewall perpendicularity and surface roughness. A three-dimensional geometric model of the high aspect ratio structure was established based on the measurement data. The corrosion rate of wafer materials by etchant under different concentrations and temperatures was tested experimentally. A correlation database of etchant concentration, temperature and corrosion rate was established, and a physicochemical property model of the etchant was constructed based on the correlation database. Design multiple sets of experiments combining different local electrowetting voltages and programmable laser scanning path parameters, record corrosion uniformity data, corrosion rate data and structural integrity detection data corresponding to each set of experiments, and establish a synergistic model of electrowetting effect and laser scanning effect using multiple regression analysis method; The three-dimensional geometric model of the high aspect ratio structure, the physicochemical property model of the corrosion liquid, and the synergistic effect model of electrowetting effect and laser scanning effect are imported into the digital twin platform. Data interaction interfaces between the models are set to complete the construction of a digital twin simulation environment dedicated to wafer wet etching.

4. The method for adaptive adjustment of etching parameters in wafer wet etching according to claim 2, characterized in that, The overall process score for each particle is calculated based on the simulation results, including: Based on the corrosion depth data in the simulation results, the ratio of the difference between the maximum and minimum corrosion depths in different regions of the high aspect ratio structure to the average corrosion depth is calculated to obtain the corrosion uniformity index. Based on the corrosion depth and time data in the simulation results, the ratio of corrosion depth per unit time to the target corrosion rate is calculated to obtain the corrosion rate index. Based on the structural morphology data in the simulation results, the number of depressions, the number of protrusions and the maximum defect size of the structural sidewalls are detected to obtain the structural integrity index. Scoring standards are established for the corrosion uniformity index, the corrosion rate index, and the structural integrity index, respectively, and each index is quantified into dimensionless corrosion uniformity index score value, corrosion rate index score value, and structural integrity index score value. According to preset weights, the corrosion uniformity index score, corrosion rate index score, and structural integrity index score are weighted and summed to calculate the particle's overall process score.

5. The method for adaptive adjustment of etching parameters in wafer wet etching according to claim 1, characterized in that, During the wet etching process, in-situ monitoring is performed on the interior of the high aspect ratio structure to obtain the gas-liquid interface radius of curvature and byproduct concentration gradient in real time, including: The gas-liquid interface inside the high aspect ratio structure was three-dimensionally scanned using a laser confocal microscopy system to obtain three-dimensional point cloud data of the gas-liquid interface. Extract the three-dimensional point cloud data of the gas-liquid interface in the target area, and fit the local quadratic surface based on the three-dimensional point cloud data of the region using the least squares method. Based on the surface parameters of the local quadratic surface, the average curvature of the target region is calculated, and the reciprocal of the average curvature is used as the radius of curvature of the gas-liquid interface. A specific wavelength of probe laser is emitted into the interior of the high aspect ratio structure using a micro-area laser-induced fluorescence spectroscopy system to excite and collect characteristic fluorescence signals generated by corrosion byproducts. The byproduct concentration gradient is calculated based on the signal intensity distribution of the characteristic fluorescence signal along the depth direction of the high aspect ratio structure.

6. The method for adaptive adjustment of etching parameters in wafer wet etching according to claim 1, characterized in that, Based on the radius of curvature of the gas-liquid interface and the concentration gradient of the byproducts, the interface transmission impedance value is calculated using a predefined transmission impedance calculation model, including: The real-time acquired gas-liquid interface curvature radius and byproduct concentration gradient are input into a predefined transmission impedance calculation model to calculate the interface transmission impedance value. The interface transmission impedance value is used to characterize the comprehensive resistance of the corrosive liquid to filling and transferring mass within the high aspect ratio structure.

7. The method for adaptive adjustment of etching parameters in wafer wet etching according to claim 1, characterized in that, The interface transmission impedance value is input into a corrosion parameter control model pre-trained based on deep reinforcement learning, and the output includes local electrowetting voltage adjustment and programmable laser scanning path parameter adjustment, including: Retrieve the pre-trained corrosion parameter control model; The real-time acquired interface transmission impedance value and the historical sequence data of the interface transmission impedance value are combined to form a characterization of the current process status. The current process state is input into the corrosion parameter control model, and the local electrowetting voltage adjustment and the programmable laser scanning path parameter adjustment are output.

8. The method for adaptive adjustment of etching parameters in wafer wet etching according to claim 1, characterized in that, The process of constructing the corrosion parameter control model includes: Based on the aforementioned digital twin simulation environment, a simulation training environment is established with the interface transmission impedance value as the state input and the local electrowetting voltage adjustment amount and the programmable laser scanning path parameter adjustment amount as the control action output. Based on the deep reinforcement learning framework, a corrosion parameter control model including a policy network and a value network is constructed. A reward function is designed with the optimization objective of minimizing the long-term accumulation and instantaneous fluctuation of the interface transmission impedance value; In the simulation training environment, the corrosion parameter control model is iteratively trained using a deep reinforcement learning algorithm until the control strategy converges, thus obtaining a pre-trained corrosion parameter control model.

9. The method for adaptive adjustment of etching parameters in wafer wet etching according to claim 1, characterized in that, Based on the local electrowetting voltage adjustment amount and the programmable laser scanning path parameter adjustment amount, adaptive adjustments are made to the initial local electrowetting voltage and the initial programmable laser scanning path parameters, including: The local electrowetting voltage adjustment is added to the initial local electrowetting voltage to obtain the updated local electrowetting voltage; The adjusted amount of the programmable laser scanning path parameters is added to the initial programmable laser scanning path parameters to obtain the updated programmable laser scanning path parameters; Based on the updated local electrowetting voltage, the electrowetting voltage applied to the bottom of the high aspect ratio structure is adjusted; The laser scanning path acting on the sidewall of the high aspect ratio structure is adjusted according to the updated programmable laser scanning path parameters. The interior of the high aspect ratio structure is continuously monitored in situ, and the interface transmission impedance value is calculated, new local electrowetting voltage adjustment amount and new programmable laser scanning path parameter adjustment amount are obtained until the wet etching process is completed.

10. A system for adaptively adjusting corrosion parameters in wet etching of wafers, characterized in that, The method for adaptively adjusting etching parameters for performing wafer wet etching according to any one of claims 1-9 includes: The process pre-optimization module is used to perform global optimization in a digital twin simulation environment based on the geometric features of the high aspect ratio structure on the wafer and the process requirements of wet etching. It calculates the optimized initial local electrowetting voltage and initial programmable laser scanning path parameters by using the particle swarm optimization algorithm in the digital twin simulation environment. The product monitoring module is used to perform in-situ monitoring of the interior of the high aspect ratio structure during the wet corrosion process, and to acquire the gas-liquid interface curvature radius and by-product concentration gradient in real time. The impedance calculation module is used to calculate the interface transmission impedance value based on the radius of curvature of the gas-liquid interface and the concentration gradient of the by-products using a predefined transmission impedance calculation model. The parameter tuning module is used to input the interface transmission impedance value into the corrosion parameter control model based on deep reinforcement learning pre-training, and output the local electrowetting voltage adjustment amount and the programmable laser scanning path parameter adjustment amount; The adjustment execution module adaptively adjusts the initial local electrowetting voltage and the initial programmable laser scanning path parameters based on the adjustment amount of the local electrowetting voltage and the adjustment amount of the programmable laser scanning path parameters.

Images